Data fusion strategies for small sample based on multi-class support vector machine
以多类支持向量机为基础的小样本信息融合策略*

This paper took multi-class support vector machine as classifier, and combined the classification results with Dempster-Shafer theory or other data fusion methods to solve the problems about small sample classification. Integrated the outputs of the multi-class support vector machines by maximal sum, Dempster-Shafer theory and the second multi-class support vector machine after Dempster-Shafer theory. Support vector machine was a machine learning algorithm fit for small sample, and Dempster-Shafer theory showed good performance about uncertain cases, so the combination of these two algorithms applied to the problems of small sample and improved the accuracy of classification. The experiment results show that the strategies can get good classification results in condition of small sample.